基于深度强化学习的5G电力虚拟专网端到端确定性切片编排算法研究
DOI:
CSTR:
作者:
作者单位:

国网福建省电力有限公司经济技术研究院

作者简介:

通讯作者:

中图分类号:

基金项目:

国网福建经研院2024年基于多类业务的5G确定性网络仿真技术研究


Research on End-to-End Deterministic Slice Orchestration Algorithm in 5G Power Virtual Private Network Based on Deep Reinforcement Learning
Author:
Affiliation:

Economic and Technological Research Institute, State Grid Fujian Electric Power Co. Ltd

Fund Project:

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    5G 电力虚拟专网中不同的电力业务对时延、可靠性等指标有不同的确定性要求,网络切片可以通过创建虚拟网络来为业务提供定制化服务。为了确保服务质量并有效编排网络切片,保障各类业务的确定性服务性能,本文对5G电力虚拟专网中的端到端确定性切片编排问题进行研究。首先,设计了一种5G电力虚拟专网和端到端网络切片相结合的架构,并建立了包含网元映射过程的5G电力虚拟专用端到端网络切片VNF部署系统模型。其次,分析业务的服务需求,提出了一种同时考虑AoI和可靠性的新型资源分配优化模型,将系统长期的平均满意度作为优化目标。最后,为解决优化问题,提出了一种基于深度强化学习的确定性业务资源分配算法进行求解。仿真结果表明,所提算法在收敛性、时延、平均AoI等方面能够实现更好性能,满足5G电力虚拟专网中业务的确定性需求。

    Abstract:

    In the 5G power virtual private network (VPN), different power services have varying deterministic requirements for latency, reliability, and other metrics. Network slicing can provide customized services for these needs by creating virtual networks. To ensure service quality and effectively orchestrate network slices, guaranteeing the deterministic service performance of various types of services, this paper studies the end-to-end deterministic slice orchestration problem in the 5G power VPN. First, a 5G power VPN architecture combined with end-to-end network slicing is designed, and an end-to-end network slice virtual network function (VNF) deployment system model is established, which includes the network element mapping process. Next, the service requirements of different power services are analyzed, and a novel resource allocation optimization model is proposed, which simultaneously considers age of information (AoI) and reliability, with the long-term average satisfaction of system as the optimization objective. Finally, to solve the optimization problem, a deterministic service resource allocation algorithm based on deep reinforcement learning is proposed. Simulation results show that the proposed algorithm achieves better performance in terms of convergence, latency, and average AoI compared to baseline algorithms, meeting the deterministic requirements of services in the 5G power VPN.

    参考文献
    相似文献
    引证文献
引用本文
分享
相关视频

文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2025-04-22
  • 最后修改日期:2025-04-22
  • 录用日期:2026-07-08
  • 在线发布日期:
  • 出版日期:
文章二维码